Machine learning researcher at Oxford working across molecular property prediction and AI-driven materials discovery
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
1
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-16) and cover the author's entire public record, not just the papers listed here. Reproducibility signal comes from Evolve's own reviews.
Research profile
Output—
The output axis is unscored because volume metrics are incomplete in this snapshot; only partial signals (8 works in the last three years) are available. This axis would measure total publications, annual counts, and productivity trend.
Impact—
The impact axis is unscored for lack of adequate citation data — the snapshot shows only an h-index of 1 and an i10-index of 0. This axis would measure citations, citation quality, and influence on downstream research.
Focus0/100
The focus dimension scores 0, the lowest reading in this snapshot. The field distribution splits evenly — 0.5 machine learning for molecular science and 0.5 AI-driven materials discovery — and OpenAlex topics scatter across at least five clusters: advanced neural network applications (6), domain adaptation and few-shot learning (6), topic modeling (4), machine learning and data classification (4), and neural networks and applications (3). Work so far is breadth-first rather than concentrated in one niche.
Collaboration36/100
Collaboration scores 36. The raw record lists 5 distinct coauthors — Moritz Friedemann, Zachary Shinnick, Xiaotang Feng, and Philip Torr among them — and both cataloged 2026 preprints are last-authored and non-corresponding. This points to a small, stable core-team pattern rather than the wide coauthor networks typical of large experimental consortia.
Activity64/100
Activity scores 64. The OpenAlex counts-by-year record shows 2 works per year from 2020 to 2022 and in 2024–2025, a gap in 2023, and an uptick to 4 works in 2026 — the highest annual count in the record — giving 8 works over the last three years. The recent acceleration supports the above-midpoint score.
Reproducibility55/100
Reproducibility scores 55, computed as the mean across 2 assessed papers. Evolve rates both cataloged preprints as medium readiness: the procedural pretraining study is described as statistically thin, and the online surrogate repair work reports overlapping confidence intervals with no code release, a limitation the review flags explicitly. Reproducibility evidence is therefore middling and thin in coverage.
Research summary
Bruno Andreis is a machine learning researcher at the University of Oxford whose OpenAlex-resolved profile sits at the interface of molecular machine learning and AI-driven materials discovery, with the snapshot's field distribution split evenly between the two. Topic clusters are led by advanced neural network applications and domain adaptation / few-shot learning (6 works each), followed by topic modeling and machine learning / data classification (4 each). Output is early-career in scale: 8 works in the last three years, 2026 the most active year (4 works), an h-index of 1 and an i10-index of 0. Two 2026 arXiv preprints appear in the Evolve catalog, both last-authored and non-corresponding: procedural pretraining with synthetic data for molecular property prediction, which cut Lipophilicity error by 4.8% even beyond molecular pretraining, and online surrogate repair for closed-loop discovery agents, which reported large consistent regret gains despite overlapping confidence intervals. Evolve rates both papers medium reproducibility readiness; the aggregate reproducibility mean is 55 across 2 assessed papers. Collaboration is compact — 5 distinct coauthors recorded — with the collaboration score at 36, consistent with small-team work rather than large consortia.
Collaboration pattern
The snapshot records 5 distinct coauthors, and the collaboration dimension scores 36 — a compact, small-team pattern rather than large-consortium work. Repeated collaborators include Moritz Friedemann, Zachary Shinnick, Xiaotang Feng, and Philip Torr, who also appears as "Philip H. S. Torr", likely a duplicate name form in the record. Both cataloged 2026 preprints are last-authored but not corresponding, consistent with a senior-position yet non-corresponding role within the group.
Research keywords
Highlights
Procedural pretraining cuts Lipophilicity error by 4.8% beyond molecular pretraining
Online surrogate repair yields large regret gains in closed-loop discovery agents
Timeline
- 2020Earliest year in the OpenAlex counts-by-year record, with 2 works attributed.
- 20222 works recorded, closing a steady three-year stretch of two publications per year (2020–2022).
- 20242 works recorded, resuming output after a gap year (no works recorded in 2023).
- 20252 works recorded, matching the previous year's pace.
- 2026Most active year with 4 works recorded, including both Evolve-cataloged arXiv preprints — procedural pretraining for molecular property prediction and online surrogate repair for closed-loop discovery — both last-authored.
Papers in the Evolve catalog
Frequent co-authors
About this profile
This profile is AI-structured by Evolve.Pub from public bibliographic data (OpenAlex) and Evolve's own paper reviews. It is not written or endorsed by the researcher unless the profile is claimed. Radar dimensions are computed from real bibliometrics with fixed formulas — raw figures are shown alongside; missing dimensions mean insufficient data, never zero. Errors? Claim the profile to correct or hide it.
- Bibliometric metrics come from an OpenAlex identity snapshot only; no ORCID was provided, so name resolution relies on public bibliographic data.
- Output and impact dimensions lack scored data; no citation counts beyond h-index 1 and i10-index 0 are available.
- Only 2 papers (both 2026 arXiv preprints) have been reviewed in the Evolve catalog, so profile depth is limited.
- The reproducibility score of 55 covers only 2 assessed papers; code-release status beyond Evolve review notes is not recorded.